fix decimal precision issue

[mvadari]

I was having issues where a Decimal object was being outputted incorrectly in the table - the value was 99999998999.999980 and tabulate was called with floatfmt=",.6f", and the output in that cell was 99,999,998,999.999985. This PR fixes that problem.

[astanin]

I suppose that tabulate should not convert all float values to Decimal, but rather stick to the basic float type instead.
However, if the user passes Decimal explicitly, it should printed as you suggest.

So the tests should be something like this:

>>> tabulate([[float("0.123456789012345678901")]], tablefmt='plain', floatfmt=".21f")
'0.123456789012345677370'
>>> tabulate([[Decimal("0.12345678901234567891")]], tablefmt='plain', floatfmt=".21f")
'0.123456789012345678901'